Clinical detection of optic nerve damage: measuring changes in cup steepness with use of a new image alignment...

The purpose of this study was to study the effect of a subpixel image alignment algorithm on the standard deviation (SD) of mean topography images obtained by laser scanning tomography and to evaluate changes of the cup shape measure parameter (CSM) over time based upon the individual parameter variability using the new algorithm. Triple measurements from optic nerve heads of 132 eyes of 132 subjects were obtained using the Heidelberg Retina Tomograph HRT. To calculate a mean topography image from three single topography images, alignment of the raw optical section image data was performed with the standard software and again with a new subpixel-based image alignment algorithm. The effect on the averaged (SD) of the mean topography images was evaluated. CSM was evaluated in 15 eyes of 15 normal subjects (N) and 28 eyes of 14 glaucoma patients (G) over a period of 28.6 +/- 4.6 months (N) and 28.56 +/- 5.2 months (G) respectively. A change in the CSM value over time was considered significant if CSM measurements exceeded two standard deviations of this variable determined for the individual eye. Mean-topography image SD was 22.86 +/- 8.2 microns (min. 9.5 microm; max. 47.8 microm) with the standard alignment procedure and 15.46 +/- 6.8 microm (min. 6.8 microm; max. 42.8 microm) with the new algorithm. The average SD improvement was 7.46 +/- 3.9 microns (min. -8.1 microm; max. 28.7 microm). The coefficient of correlation of both methods was R(2) = 0.77 (p < 0.0001). No control group eye demonstrated significant changes of CSM in the follow-up period. The CSM indicated an increase in cup steepness in 4 eyes of 4 glaucoma patients. In one of these four eyes, a deterioration of the visual field was identified by white on white perimetry. The new image alignment algorithm significantly reduces the SD of mean topography images calculated from identical raw data. If topometric variables are evaluated over time, the individual variability of data should be taken into account.